Congratulazioni!

CI/CD per Machine Learning

Ravi Bhadauria

Machine Learning Engineer

Sintassi YAML

• Rientro
• Mappature a: 1
• Array
  • In linea [1, 2]
  • A blocchi

    - 1
    - 2
    

• Stringhe multi-riga
  • Indicatori di stile (|, >)
  • Indicatori di troncamento (-, +)

GitHub Actions

• Workflow (pipeline)
• Eventi (on)
• Job (jobs)
• Runner (runs-on)
• Step (steps)
• Contesti
• Secret e variabili d'ambiente (GITHUB_TOKEN)
• Action (checkout, setup-python)
  • CML: cml comment create, cml pr create

Versionare i dati e creare pipeline riproducibili

• dvc init
• Remote DVC (anche locali)
• dvc push, dvc pull
• dvc repro <target>
• dvc metrics show/diff
• dvc plots show/diff

• YAML di DVC (dvc.yaml)
  • Step o target
  • Comandi (cmd)
  • Dipendenze (deps)
  • Output (outs)
  • Metriche (metrics)
  • Grafici (plots)

Risorse DataCamp

• Corsi

  • Developing Machine Learning Models for Production
  • MLOps Deployment and Life Cycling
  • Fully Automated MLOps
  • Introduction to DevOps

• Blog

  • Version Control For Data Science

Approfondimenti

CI/CD e modelli di branching

DevOps vs MLOps

Data Version Control

Model registry e gestione del ciclo di vita

Grazie!

CI/CD per Machine Learning